Independent evaluation of a multi-view multi-task convolutional neural network breast cancer classification model using Finnish mammography screening data.

BACKGROUND: Development of deep convolutional neural networks for breast cancer classification has taken significant steps towards clinical adoption. It is though unclear how the models perform for unseen data, and what is required to adapt them to different demographic populations. In this retrospective study, we adopt an openly available pre-trained mammography breast cancer multi-view classification model and evaluate it by utilizing an independent Finnish dataset. METHODS: Transfer learning was used, and the pre-trained model was finetuned with 8,829 examinations from the Finnish dataset (4,321 normal, 362 malignant and 4,146 benign examinations). Holdout dataset with 2,208 examinations from the Finnish dataset (1,082 normal, 70 malignant and 1,056 benign examinations) was used in the evaluation. The performance was also evaluated on a manually annotated malignant suspect subset. Receiver Operating Characteristic (ROC) and Precision-Recall curves were used to performance measures. RESULTS: The Area Under ROC [95%CI] values for malignancy classification obtained with the finetuned model for the entire holdout set were 0.82 [0.76, 0.87], 0.84 [0.77, 0.89], 0.85 [0.79, 0.90], and 0.83 [0.76, 0.89] for R-MLO, L-MLO, R-CC and L-CC views respectively. Performance on the malignant suspect subset was slightly better. On the auxiliary benign classification task performance remained low. CONCLUSIONS: The results indicate that the model performs well also in an out-of-distribution setting. Finetuning allowed the model to adapt to some of the underlying local demographics. Future research should concentrate to identify breast cancer subgroups adversely affecting performance, as it is a requirement for increasing the model's readiness level for a clinical setting.

Multimodality scoring of chondral injuries in the equine fetlock joint ex vivo.

OBJECTIVE: We investigate the potential of a prototype multimodality arthroscope, combining ultrasound, optical coherence tomography (OCT) and arthroscopic indentation device, for assessing cartilage lesions, and compare the reliability of this approach with conventional arthroscopic scoring ex vivo. DESIGN: Areas of interest (AIs, N = 43) were selected from equine fetlock joints (N = 5). Blind-coded AIs were independently scored by two equine surgeons employing International Cartilage Repair Society (ICRS) scoring system via conventional arthroscope and multimodality arthroscope, in which high-frequency ultrasound and OCT catheters were attached to an arthroscopic indentation device. In addition, cartilage stiffness was measured with the indentation device, and lesions in OCT images scored using custom-made automated software. Measurements and scorings were performed twice in two separate rounds. Finally, the scores were compared to histological ICRS scores. RESULTS: OCT and arthroscopic examinations showed the highest average agreements (55.2%) between the scoring by surgeons and histology scores, whereas ultrasound had the lowest (50.6%). Average intraobserver agreements of surgeons and interobserver agreements between rounds were, respectively, for conventional arthroscope (68.6%, 69.8%), ultrasound (68.6%, 68.6%), OCT (65.1%, 61.7%) and automated software (65.1%, 59.3%). CONCLUSIONS: OCT imaging supplemented with the automated software provided the most reliable lesion scoring. However, limited penetration depth of light limits the clinical potential of OCT in assessing human cartilage thickness; thus, the combination of OCT and ultrasound could be optimal for reliable diagnostics. Present findings suggest imaging and quantitatively analyzing the entire articular surface to eliminate surgeon-related variation in the selection of the most severe lesion to be scored.